Pull Request Explorer

Summary

Fix QStash minimum settling delay enforcement

Ensures reliable queue dispatch timing by enforcing a minimum local delay for sub‑second QStash delays, preventing immediate delivery collapse and improving retry reliability.

Health Assessment

• • Fast cycle time, single review, minimal changes, low risk.

Summary

refactor: move tool executors into agent runtime

Reorganizes tool execution logic into the agent-runtime package, adding server adapters and transport extensions for improved modularity and testability.

Health Assessment

• • Fast review and merge with minimal comments suggest confidence in the refactor, despite the large code movement.

Summary

Add device share support and workspace optimization

Enables users to share devices and improves workspace performance, enhancing collaboration.

Health Assessment

• • PR merged in 16 hours with minimal review, indicating straightforward configuration changes. The large line count reflects extensive locale updates but no code logic changes.

Summary

Fix chat message opacity

Prevents user messages from appearing dimmed while being sent, improving the chat experience.

Health Assessment

• • The PR was reviewed and merged quickly, indicating a smooth development process.

Summary

Add Grok 4.5 reasoning effort parameter

Adds a UI control for Grok 4.5 reasoning effort, enabling users to fine‑tune model behavior and improve response quality.

Health Assessment

• • Rapid review and merge with minimal changes indicates a straightforward, low‑risk enhancement.

Summary

Fix late hetero terminal event handling

Ensures heterogeneous agent sessions correctly wait for all terminal events before finalizing, preventing premature completion and data loss.

Health Assessment

• • Quick fix with minimal rework; AI review comments helped identify edge cases and validate the solution.

Summary

Health Assessment

Summary

Fix ChatInput warning collapsing actions

Prevents warning messages from causing UI collapse in chat input, improving user experience on desktop and mobile.

Health Assessment

• • PR merged in 1.7 hours with minimal rework; AI review tools provided feedback but no code changes were required.

Summary

Polish agent-browser builtin skill prompt and description

Refines the agent-browser skill by simplifying prompts, adding command examples, and tightening description, improving reliability and usability.

Health Assessment

• • Quick polish with minimal changes, reviewed by AI, merged immediately.

Summary

Fix sync of agent git metadata from CLI

Ensures the agent accurately tracks git worktree and branch changes, improving consistency across CLI interactions.

Health Assessment

• • Quick turnaround with AI-assisted review, minimal rework, and a focused bugfix.

Summary

Fix abandoned operation lifecycle finalization

Ensures agent runtime correctly finalizes abandoned operations, preventing stuck states and dispatching completion hooks.

Health Assessment

• • Merged within 6 hours with minimal rework; AI review provided suggestions and the change was straightforward to implement.

Summary

Reconcile gateway connections with database

Ensures stale bot connections are removed and only active providers remain, preventing unnecessary polling and message delivery for disabled or misconfigured bots, thereby improving reliability and resource efficiency.

Health Assessment

• • Merged within 2.7 hours with no human review friction; AI review bots provided automated feedback, leading to a quick, low-risk fix.

Summary

Improve input loading responsiveness for actions

This change makes the input field show a loading state instantly when users trigger approve, retry, or tool interactions, reducing perceived latency and improving the chat experience.

Health Assessment

• • Fast review cycle and immediate merge, but multiple commits indicate iterative refinement; performance improvement enhances user experience.

Summary

Unblock onboarding startup on desktop

Fixes desktop onboarding startup by allowing the cache hydration gate to release without waiting for a user ID, improving the initial user experience.

Health Assessment

• • Quick, single‑commit fix with AI‑assisted review, resulting in a fast cycle time and minimal scope.

Summary

Polish WeChat messaging announcement

Updates the WeChat messaging announcement to improve clarity and localization, ensuring users understand new paid plan requirements and time zones.

Health Assessment

• • Quick merge with minimal changes and no review friction.

Summary

Fix web execution target display and working directory leak

Corrects UI mislabeling of execution targets on web and prevents leaking of working directory paths into prompts, improving user experience and security.

Health Assessment

• • Fast review and merge with minimal rework; AI review bots provided suggestions but no human rework was needed.

Summary

Add personal credential sharing to workspace

Enables workspace members to share personal credentials into the workspace organization, improving collaboration and AI context accuracy.

Health Assessment

• • PR merged quickly after review, but required multiple commits to address review feedback, indicating moderate iteration but overall fast cycle.

Summary

Add WeChat chat adjustment announcement

Publish a product announcement on the website changelog about upcoming changes to WeChat chat requiring a subscription plan.

Health Assessment

• • Quickly merged after a single review, indicating a straightforward documentation update.

Summary

Fix multi-select visibility for hetero messages

Restores the multi-select action in hetero-agent menus, ensuring users can batch-select messages for forwarding or deletion. This improves usability by keeping essential actions available while trimming unsupported ones.

Health Assessment

• • Quick review and merge within an hour, indicating straightforward change with minimal friction.

Summary

Add task_topics.trigger and user_connectors.agent_id

Consolidates database schema changes from two branches into a single migration, preventing conflicts and enabling smoother integration.

Health Assessment

• • Large migration combined from two branches; review was quick but overall cycle time was long due to the size of the changes, indicating moderate to high risk of schema conflicts.

Summary

Fix agent-runtime to persist empty completion retry diagnostics

Improves observability and post-deploy triage for ModelEmptyCompletion errors by preserving retry attempt history in agent-runtime trace snapshots.

Health Assessment

• • The PR was reviewed and merged quickly, indicating low review friction.
• • The changes are focused on improving error diagnostics, which is a relatively low-risk change.

Summary

Fix credential routing to organization credentials

Ensures workspace agent operations use shared organization credentials instead of the operator’s personal account, improving security and consistency for workspace users.

Health Assessment

• • The PR was reviewed and merged within an hour, indicating a straightforward bug fix with minimal friction and low risk.

Summary

Fix agent browser lifecycle and install prompts

Improves agent safety by ensuring browsers close properly and prompting users before installing system software, reducing accidental user interference.

Health Assessment

• • Quick turnaround with minimal changes and no back‑and‑forth, indicating a straightforward bug fix.

Summary

Align review gutter button style

UI styling adjustment to match GitHub review button, improving visual consistency.

Health Assessment

• • Rapid review and merge with minimal changes, indicating low complexity and high confidence.

Summary

Batch hetero message writes for performance

Improves message persistence by batching writes, reducing latency and preventing race conditions, enhancing user experience during agent interactions.

Health Assessment

• • Fast review and merge with minimal back‑and‑forth, but large code change indicates significant refactor of message handling.

Summary

Fix chat text selection lab gating

Restores default chat performance by gating the text selection layer and enables task verify by default, improving user experience.

Health Assessment

• • Quick turnaround with AI review, minimal rework, fast merge.

Summary

Fix connect service env scoping

Ensures the connect service uses a dedicated environment file, preventing global environment pollution while preserving user secrets such as API keys.

Health Assessment

• • The PR was reviewed immediately and required only a single commit, indicating a smooth review process. The moderate code change size and quick cycle time suggest low risk and efficient implementation.

Summary

Refactor LLM error classifier

Move LLM error classifier into agent-runtime for better organization and maintainability

Health Assessment

• • The PR was merged quickly, but the size of the changes is large, indicating potential complexity

Summary

Fix first paint blocking until userId resolves

This change stops the app from rendering under an anonymous cache scope, preventing stale data leaks. It ensures a consistent user experience by waiting for user identity resolution before displaying content.

Health Assessment

• • Fast review and merge with minimal rework, quickly addressing a critical caching bug that could cause stale data and orphaned cache entries.

Summary

Remove legacy storage clear data flow

Eliminates obsolete storage clearing controls and associated services, tests, and locale keys, simplifying storage management and reducing code complexity.

Health Assessment

• • The PR was reviewed and merged within 10 hours, with a single commit and no back‑and‑forth, indicating a smooth, low‑friction process. The removal of legacy code reduces technical debt and streamlines the storage feature.

Summary

Keep agent-testing living log generic

Simplifies documentation by removing environment‑specific notes, improving clarity and maintainability for developers.

Health Assessment

• • Fast merge with minimal changes and no tests; low risk.

Summary

Fix workspace org-not-found handling in CredsList

Adds defensive handling for missing workspace org, showing setup guidance instead of error, improving user experience.

Health Assessment

• • PR received immediate AI‑review feedback but required multiple commits to resolve conflicts and refine logic, indicating a complex integration with workspace credentials.

Summary

Fix home refresh load reduction

Improves performance by reducing polling frequency and optimizing data ordering, lowering server load and enhancing user experience.

Health Assessment

• • Quick, low‑impact fix with minimal rework; AI review provided automated feedback and accelerated approval.

Summary

Fix sidebar sorting to match server order

Aligns client‑side topic ordering with server activity timestamps, eliminating list jumps and improving user experience.

Health Assessment

Summary

Fix agent disappearance after workspace transfer

Ensures agents remain visible after workspace transfer, preventing data loss and improving user experience.

Health Assessment

• • The PR was reviewed and merged within 13 hours, indicating a smooth process.
• • AI assistance helped produce a concise fix with minimal rework.

Summary

Fix error classification and surface PG causes

Improves error handling by adding missing error patterns and exposing database causes, reducing debugging time and improving reliability.

Health Assessment

• • Quick turnaround with minimal rework; AI review bot flagged rate limits but did not block merge.

Summary

Optimize workspace features and privacy

Adds privacy controls for agents and tasks, improves avatar i18n, and enhances agent testing, reducing errors and improving user experience.

Health Assessment

• • Rapid 2.5‑hour cycle with AI‑assisted review and multiple iterations, yet final merge indicates efficient resolution of complex changes.

Summary

Add locale support for notifications

Enables notification messages in multiple languages, improving user experience for international users.

Health Assessment

• • Fast cycle time with minimal review, but large number of added lines indicates extensive localization work.

Summary

Add quiet message batch mutation

Adds a batch mutation endpoint for messages, reducing database round‑trips and improving streaming client performance.

Health Assessment

• • Fast cycle time and minimal rework indicate a smooth review process; AI review comments helped surface issues early.

Summary

Fix error pattern classification in model runtime

Improves error handling by adding safe error patterns and comprehensive tests, reducing false positives and enhancing user experience.

Health Assessment

• • Fast cycle time with minimal rework; AI review bots provided feedback but no code changes were required.

Summary

Add message text selection actions

Users can now quote rendered Markdown/text directly into the conversation, streamlining content reuse and improving workflow.

Health Assessment

• • Fast turnaround with minimal review friction; PR merged within 9 hours after a single review comment.

Summary

Fix embedding model list in service selector

Adds an embedding‑specific model list to the memory selector, improving accuracy and user experience.

Health Assessment

• • Quick fix with minimal rework, fast review, no blockers.

Summary

Remove sandbox input notice from chat input

The temporary Cloud Sandbox notice is removed, restoring a clean chat input area. This fixes layout issues and keeps the existing model-unavailable warning.

Health Assessment

• • Quick, single-commit fix with no review friction; minimal scope and low risk to production.

Summary

Refactor agent-runtime: host resolve tools and retry helpers

This refactor moves tool resolution and retry logic into the agent-runtime package, decoupling server-specific behavior and enabling reuse across deployments. It simplifies maintenance and enhances consistency across environments.

Health Assessment

• • Fast cycle time and minimal review rounds indicate a low-risk refactor with high confidence.

Summary

Fix gateway live text rendering

Ensures gateway agent responses render immediately by inserting a message shell on stream start, eliminating delayed pop‑in and improving user experience.

Health Assessment

• • Fast turnaround with immediate review and minimal rework; the PR resolved a UI rendering issue quickly.

Summary

Fix routing for Codex new topic streams

Resolves issue where Codex stream events were not displayed in the UI due to incorrect routing.

Health Assessment

• • The PR was reviewed and merged quickly, with a cycle time of 12.8 hours.

Summary

Fix hetero: persist resume session id at stream_start

Ensures resume session IDs are persisted even when errors occur, preventing loss of context in heterogeneous agent sessions.

Health Assessment

• • Fast review and merge within 1.4 hours, minimal changes, quick fix to prevent context loss.

Summary

Fix restore-to-input for heterogeneous agents

Adds the ‘Restore to Input’ action to the right‑click menu for Claude Code/Codex agents, enabling users to quickly re‑send prompts and attachments after errors or context loss.

Health Assessment

• • The PR was reviewed and merged within 1.5 hours with minimal changes, indicating a smooth, low‑risk process.

Summary

Fix worktree directory label and removal toast

Ensures the worktree directory label reflects the source repository path and adds a loading toast during worktree removal, improving user clarity and feedback.

Health Assessment

• • Merged within 45 minutes with minimal rework, quick review, and straightforward bug fixes.

Summary

Refresh topic PR status on detail

Adds functionality to refresh GitHub PR metadata when viewing a topic detail, ensuring the UI shows up-to-date PR status without affecting topic state.

Health Assessment

• • The PR was merged in under 12 hours with only two AI‑driven review comments, indicating a smooth review process. The large code change was completed quickly, suggesting the team was well‑aligned on the feature scope.